Sports Video Classification Framework Using Enhanced Threshold Based Keyframe Selection Algorithm and Customized CNN on UCF101 and Sports1-M Dataset

Table 4

Experimental setup [24].

Hyper-parameters	Values
Hardware resource	Single CPU
Number of classes	10(Basketball, diving, golf, horse riding, kicking-front, running, skate boarding-front, swing-bench, walk-front, weight lifting)
Image/Frame size	227 × 227
Output size in fully connected layer	10 -no. of classes
Optimizer	SGDM - stochastic gradient descent with momentum
Epochs	6
Iterations per epoch	97
Maximum iterations	582
MiniBatch size	10
Initial learning rate	0.0001
Validation frequency	3
Kernel	3 × 3 filter
Stride	1
Activation unit	ReLU - rectified linear unit
Dropout	0.5